CN105260731A - A system and method for human face liveness detection based on light pulses - Google Patents

Abstract

The embodiment of the invention discloses a human face living body detection system and a method based on optical pulses, belonging to the field of image recognition, wherein the system comprises: the light emitting device emits light signals with different signal intensities to the object to be detected at a preset frequency; the camera device is used for acquiring a plurality of images of the object to be detected generated under optical signals with different signal intensities; and a detection device that detects whether the object to be measured is a living body based on the plurality of images. By applying the scheme provided by the embodiment of the invention to identify the target area, the situation that other people pretend to be the person passing through the target area identification by using non-living objects such as photos, videos and the like can be avoided.

Description

A system and method for human face liveness detection based on light pulses

technical field

The present invention relates to the technical field of image recognition, in particular to a system and method for human face liveness detection based on light pulses.

Background technique

Face detection and recognition technology is widely used in identity recognition and verification systems due to its advantages of non-contact and non-invasiveness (no user perception). At present, most systems use image-based face recognition technology. Due to the lack of judgment on the source of the image, it is impossible to determine whether the current image collection comes from a real person, or a still image containing a face, or a pre-recorded video. , or face masks. These image-based face recognition systems lack the link of liveness detection and cannot distinguish between face materials and real people, so there are potential dangers such as deception.

In order to overcome such problems, some systems in the prior art introduce a live body detection link. These systems can be divided into two categories: texture-based, optical flow-based, and interaction-based liveness detection. The texture-based method takes advantage of the nuances of the real face skin and the texture of the paper and display screen, and trains the detection model through texture descriptors and machine learning models, but this method is easily affected by the environment (such as lighting, etc.) , and overfitting to the training samples, the generalization is not good; the method based on optical flow obtains the optical flow characteristics of different three-dimensional structures of real faces, paper and display screens during motion by establishing an optical flow field model. The disadvantages are The user is required to have a certain movement, and the result is affected by the accuracy of the optical flow algorithm; the interaction-based method obtains relevant feedback during the interaction with the user to determine whether it is a real person. For example, the system prompts the detected object to complete specified actions, such as turning head, blinking, etc., or simple questions and answers. Pre-produced material (still pictures or videos) cannot predict the interactive actions to be performed and cannot provide real-time feedback, based on which it can be judged to be non-living. The interactive method is an intrusive and non-silent detection method that requires user-assisted operations, which increases the user's additional usage burden. For some access control detection systems that come in and out frequently, a large number of repetitive operations reduce the user experience.

Contents of the invention

The embodiment of the present invention discloses a human face living body detection system and method based on light pulses, which can accurately determine whether the object to be detected is a living body, and simplifies the operation process compared with the traditional living body identification method.

In order to achieve the above purpose, the embodiment of the present invention discloses a human face liveness detection system based on light pulses, the system includes:

A light-emitting device, the light-emitting device emits light signals of different signal intensities to the object to be measured at a preset frequency;

an imaging device, the imaging device is used to acquire multiple images of the object to be measured under light signals of different signal intensities; and

A detection device, the detection device detects whether the object to be detected is a living body based on the plurality of images.

As a specific implementation manner of the embodiment of the present invention, the detection device includes:

an extracting device, the extracting device extracts target areas of the object to be measured on multiple images, and acquires feature point information of the target areas; and

A judging device, which judges whether the object to be measured is a living body based on the target area and the feature point information.

As a specific implementation of the embodiments of the present invention, the extraction device is also used for:

calculating the variance between the target area and the feature points on the current image acquired by the camera;

When the variance is lower than the first threshold, the current image is discarded, and the image of the object to be measured is acquired from the camera device again.

As a specific implementation of the embodiments of the present invention, the extraction device is also used for:

calculating the average brightness value of the target area on each image captured by the camera;

Discarding is performed on images whose average brightness value is greater than the second threshold or lower than the third threshold.

As a specific implementation of the embodiments of the present invention, the extraction device is also used for:

Within the range of preset brightness values, the image with the lowest average brightness value of key points is selected as the first image, and the image with the highest average brightness value of key points is selected as the second image.

As a specific implementation manner of the embodiment of the present invention, the detection device further includes:

A preprocessing device, configured to perform image preprocessing on the first image and the second image.

As a specific implementation manner of the embodiment of the present invention, the detecting whether the object to be tested is a living body based on the plurality of images includes:

The target areas of the first image and the second image are respectively subjected to normalization processing, and the feature points are mapped to a unified standard position.

As a specific implementation manner of the embodiment of the present invention, performing normalization processing on the target area and mapping the feature points to a unified standard position includes:

The transformation matrices between the key points of the target area and the predetermined average target area on the first image and the second image are respectively calculated, and the key points of the target area are mapped to standard positions.

As a specific implementation manner of the embodiment of the present invention, the detecting whether the object to be tested is a living body based on the plurality of images further includes:

Difference processing is performed on the normalized first image and the second image to obtain a difference map of the target area.

As a specific implementation manner of the embodiment of the present invention, the detecting whether the object to be tested is a living body based on the plurality of images further includes:

Extract discrete Fourier transform features in a difference map:

$\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; \&pi;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; ln</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; i</span>}}$

Where e ^ix =cosx+sinx, I(m,n) is the pixel value of the difference map of the target area, and N is the size of the difference map.

As a specific implementation manner of the embodiment of the present invention, the detecting whether the object to be tested is a living body based on the plurality of images further includes:

obtain the graphics type of the first image and the second image; and

Auxiliary features of the first image and the second image are extracted based on the image type.

As a specific implementation manner of the embodiment of the present invention, the detecting whether the object to be tested is a living body based on the plurality of images further includes:

constructing a feature vector using the discrete Fourier transform feature and the auxiliary feature;

The feature vector is used as an input, and is analyzed by a classifier based on reflective characteristics established by machine learning to determine whether the target to be tested is a living body.

In order to achieve the above purpose, the embodiment of the present invention also discloses a method for detecting human face liveness based on light pulses, which is characterized in that the method includes:

Transmit optical signals with different signal strengths to the object to be measured at a preset frequency;

acquiring a plurality of images of the object to be measured under light signals of different signal intensities; and

Based on the plurality of images, it is detected whether the object to be detected is a living body.

As a specific implementation manner of the embodiment of the present invention, the detecting whether the object to be tested is a living body based on the plurality of images includes:

Extracting target areas of the object to be measured on multiple images, and acquiring feature point information of the target areas; and

Based on the target area and the feature point information, it is judged whether the object to be measured is a living body.

As a specific implementation manner of the embodiment of the present invention, the acquisition of multiple images of the object to be measured under optical signals of different signal intensities includes:

calculating the variance between the target area and the feature points on the current image acquired by the camera;

When the variance is lower than the first threshold, the current image is discarded, and the image of the object to be measured is acquired from the camera device again.

As a specific implementation manner of the embodiment of the present invention, the acquisition of multiple images of the object to be measured under optical signals of different signal intensities further includes:

calculating the average brightness value of the target area on each image captured by the camera;

Discarding is performed on images whose average brightness value is greater than the second threshold or lower than the third threshold.

As a specific implementation manner of the embodiment of the present invention, the acquisition of multiple images of the object to be measured under optical signals of different signal intensities further includes:

Within the range of preset brightness values, the image with the lowest average brightness value of key points is selected as the first image, and the image with the highest average brightness value of key points is selected as the second image.

As a specific implementation manner of the embodiment of the present invention, the method further includes:

Image preprocessing is performed on the first image and the second image.

As a specific implementation manner of the embodiment of the present invention, the detecting whether the object to be tested is a living body based on the plurality of images includes:

The target areas of the first image and the second image are respectively subjected to normalization processing, and the feature points are mapped to a unified standard position.

As a specific implementation manner of the embodiment of the present invention, performing normalization processing on the target area and mapping the feature points to a unified standard position includes:

The transformation matrices between the key points of the target area and the predetermined average target area on the first image and the second image are respectively calculated, and the key points of the target area are mapped to standard positions.

As a specific implementation manner of the embodiment of the present invention, the detecting whether the object to be tested is a living body based on the plurality of images further includes:

Difference processing is performed on the normalized first image and the second image to obtain a difference map of the target area.

As a specific implementation manner of the embodiment of the present invention, the detecting whether the object to be tested is a living body based on the plurality of images further includes:

Extract discrete Fourier transform features in a difference map:

$\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; \&pi;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; ln</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; i</span>}}$

Where e ^ix =cosx+sinx, I(m,n) is the pixel value of the difference map of the target area, and N is the size of the difference map.

As a specific implementation manner of the embodiment of the present invention, the detecting whether the object to be tested is a living body based on the plurality of images further includes:

obtain the graphics type of the first image and the second image; and

Auxiliary features of the first image and the second image are extracted based on the image type.

As a specific implementation manner of the embodiment of the present invention, the detecting whether the object to be tested is a living body based on the plurality of images further includes:

constructing a feature vector using the discrete Fourier transform feature and the auxiliary feature;

The feature vector is used as an input, and is analyzed by a classifier based on reflective characteristics established by machine learning to determine whether the target to be tested is a living body.

It can be seen from the above that in the solutions provided by the embodiments of the present invention, irregular flash pulses are emitted continuously, and the reflections of the human face, such as changes in skin gloss, are collected in real time. Due to the relatively uniform reflectivity of pictures, video playback screens, and masks, the reflective properties of real faces show uneven distribution due to non-uniform skin and surface fluctuations. Determines whether the current face is from a real person or from pre-produced footage by comparing the change in the reflection of the face under flashing and non-flashing conditions. As many of the current portable devices have their own light source emission functions such as shooting cameras and flashlights, the required functions can be met through the flashing functions of the shooting cameras and flashlights, and no additional equipment is required, which reduces the complexity of the system and provides friendly user experience. At the same time, in order to be able to obtain the light reflection distribution of different types of faces, and to accurately distinguish the difference between face reflections and non-face reflections such as still pictures, video playback screens, and masks, this patent also uses machine learning technology, through A large number of face and non-face training materials (including still pictures, hand-made face shape pictures, face videos, and face masks), the establishment of learning classifiers for faces and non-faces, overcoming other current methods Existing problems, so as to accurately determine various types of living human faces or non-living materials.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments of the present invention. Those skilled in the art can also obtain other drawings based on these drawings without creative work;

FIG. 1 is a structural diagram of another optical pulse-based human face liveness detection system provided by an embodiment of the present invention;

FIG. 2 is a structural diagram of a detection device for a human face liveness detection system based on light pulses provided by an embodiment of the present invention;

FIG. 3 is a structural diagram of another optical pulse-based human face liveness detection system detection device provided by an embodiment of the present invention;

FIG. 4 is a flow chart of a method for detecting human face liveness based on light pulses according to an embodiment of the present invention.

detailed description

Reference will now be made in detail to the embodiments, examples of which are illustrated in accompanying drawings 2-3. In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the present invention. It will be apparent, however, to one skilled in the art that the present invention may be practiced without these specific details. In other instances, well-known methods, procedures, components, circuits and networks have not been described in detail so as not to unnecessarily obscure aspects of the embodiments.

It should also be understood that, although the terms first, second etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, a first posture could be termed a second posture, and, similarly, a second posture could be termed a first posture, without departing from the scope of the present invention.

Here, terms used in the description of the present invention are only for describing specific embodiments, and are not intended to limit the present invention. As used in the description of the present invention and the appended claims, the singular forms "a", "an" and "the" are intended to include the plural forms as well unless the context clearly dictates otherwise. It should also be understood that the term "and/or" as used herein refers to and includes any and all possible combinations of one or more of the associated listed items. It should also be understood that the term "comprises" and/or "comprises" when used in this specification means the presence of the stated features, integers, steps, operations, elements, and/or components, but does not exclude the presence or addition of One or more other features, integers, steps, operations, elements, components, and/or collections thereof.

FIG. 1 is a schematic structural diagram of a human face liveness detection system based on light pulses provided by an embodiment of the present invention. The system 100 includes:

A light emitting device 101, the light emitting device emits light signals with different signal intensities to the object to be measured at a preset frequency.

Specifically, the light emitting device may be an independent controllable light source, or a light source integrated into the device. Continuous frames of ordinary images illuminated by a controllable light source collected by the light-emitting device in a short distance. This system uses 20 consecutive frames within one second.

In addition to visible light, light-emitting devices can generate other imaging rays such as infrared light.

An imaging device 102, configured to acquire multiple images of the object to be measured under light signals of different signal intensities.

Specifically, ordinary cameras can be used to acquire ordinary images, and ordinary images can include but not limited to grayscale images, color images or infrared images, and sources of the camera device 102 include but not limited to network cameras, surveillance cameras, and cameras that come with mobile phones.

A detecting device 103, the detecting device detects whether the object to be tested is a living body based on the plurality of images.

Specifically, the detection device 103 selects several images with better imaging quality and higher contrast from the multiple photos containing faces taken by the camera device 102. Since the reflection of a live face is more divergent than that of a photo, the , it can be judged whether the object to be tested is a living body by comparing the difference of the imaging luminance of the above photos.

As an implementation manner, the detecting device 103 further includes an extracting device 1031 .

The extracting device 1031 performs face detection on all consecutive frames, and obtains the areas where faces appear and the key points of faces. Calculate the variance of the area and key points, if it is lower than a certain threshold, it will be discarded and the liveness detection will be performed again, otherwise, an average face area will be calculated.

The extracting means 1031 is also used to calculate the average brightness in the average face area in each frame. Frames with average brightness above the upper threshold or below the lower threshold are discarded. Select the two frames with the lowest value and the highest value of the average brightness of key point changes in a certain threshold range, which are recorded as I ₁ and I ₂ .

As an optional implementation manner, the detection device 103 further includes a preprocessing device 1033 , which includes but is not limited to scaling, cropping, denoising, blurring, and the like on the two frames of light and dark images. This preprocessing operation can greatly improve the accuracy of the final judgment.

As an optional implementation, the normalization process is performed on the recognized faces. The size of the face in the face area is not fixed, so it needs to be normalized. The method includes the following steps: The key points of the face in I ₁ , and calculate the transformation matrix with the predetermined average face, map the key points of the face to the standard position, and transform the face image to the standard size. The same transformation is performed on _I2 using the matrix above.

As an optional implementation manner, after the normalized high-brightness and low-brightness images are obtained, differential processing is performed to obtain a light-dark face difference map.

Specifically, the discrete Fourier transform features are extracted in the difference map:

$\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; \&pi;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; ln</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; i</span>}}$

Where e ^ix =cosx+sinx, I(m,n) is the pixel value of the difference map of the target area, and N is the size of the difference map.

In addition to discrete Fourier transform features, other applicable features include but are not limited to: extracting pixel values for the entire face or near certain key points, LBP, Gabor, gradient histogram conversion, SIFT, SURF, etc.

The difference map can well reflect the material information of the image and the information of local illumination changes over time, but it is highly sensitive to spatial changes, so other features that are not related to spatial changes need to be extracted. If it is a color image, more auxiliary features are extracted in channels that can reflect illumination information. The channels include the V channel of HSV, the Y channel of YUV, the L channel of LAB and the weighted average of the three passes of RGB; if it is a grayscale or infrared image, use its single pass value. In this example, the Fourier transform features are extracted from the selected channels of the face area of the two frames of light and dark images, and other suitable features mentioned above may also be used.

As an optional implementation manner, a feature vector is obtained by concatenating the discrete Fourier feature and the auxiliary feature. Using the model of machine learning to build a random forest classifier for living and non-living. The model of this system uses a large number of data (500,000 groups) of human faces and non-human faces collected in different environments (such as human faces on the screen of mobile phones or tablet computers, and photos printed with different materials) for training. The input of the training algorithm is the above feature vector and expected output. Other suitable machine learning algorithms include, but are not limited to, artificial neural networks, decision trees, support vector machines, convolutional neural networks, and the like. Use the score of the classifier to make a judgment. If it is greater than or equal to the threshold, it will be judged as a living body, and if it is less than the threshold, it will be judged as a non-living body.

Fig. 4 is the flow chart of the human face detection method based on light pulse provided by the embodiment of the present invention, the method includes:

S401. Transmit optical signals with different signal intensities to the object to be measured at a preset frequency.

Specifically, the above step S501 is performed by using a light emitting device. The light emitting device may be an independent controllable light source, or a light source integrated in the device. Continuous frames of ordinary images illuminated by a controllable light source collected by the light-emitting device in a short distance. This system uses 20 consecutive frames within one second.

In addition to visible light, light-emitting devices can generate other imaging rays such as infrared light.

S402. Acquire multiple images of the object to be measured under light signals of different signal intensities.

Specifically, ordinary cameras can be used to obtain ordinary images, and ordinary images can include but not limited to grayscale images, color images or infrared images, and sources of camera devices include but not limited to network cameras, surveillance cameras, and cameras that come with mobile phones.

S403. Based on the multiple images, detect whether the object to be tested is a living body.

Specifically, several images with better imaging quality and higher contrast are selected from multiple photos containing faces taken by the camera device. Since the reflection of living human faces is more divergent than that of photos, by comparing the above Whether the object to be tested is a living body can be judged by the difference value of the imaging luminance of the photo.

As an implementation manner, face detection is performed on all consecutive frames to obtain areas where faces appear and key points of faces. Calculate the variance of the area and key points, if it is lower than a certain threshold, it will be discarded and the liveness detection will be performed again, otherwise, an average face area will be calculated. Computes the average brightness in the average face region within each frame. Frames with average brightness above the upper threshold or below the lower threshold are discarded. Select the two frames with the lowest value and the highest value of the average brightness of key point changes in a certain threshold range, which are recorded as I ₁ and I ₂ .

As an optional implementation manner, the method further includes a preprocessing step, including but not limited to scaling, cropping, denoising, blurring, etc., on the two frames of light and dark images. This preprocessing operation can greatly improve the accuracy of the final judgment.

As an optional implementation, the normalization process is performed on the recognized faces. The size of the face in the face area is not fixed, so it needs to be normalized. The method includes the following steps: The key points of the face in I ₁ , and calculate the transformation matrix with the predetermined average face, map the key points of the face to the standard position, and transform the face image to the standard size. The same transformation is performed on _I2 using the matrix above.

As an optional implementation manner, after the normalized high-brightness and low-brightness images are obtained, differential processing is performed to obtain a light-dark face difference map.

Specifically, the discrete Fourier transform features are extracted in the difference map:

$\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; \&pi;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; ln</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; i</span>}}$

Where e ^ix =cosx+sinx, I(m,n) is the pixel value of the difference map of the target area, and N is the size of the difference map.

In addition to discrete Fourier transform features, other applicable features include but are not limited to: extracting pixel values for the entire face or near certain key points, LBP, Gabor, gradient histogram conversion, SIFT, SURF, etc.

The difference map can well reflect the material information of the image and the information of local illumination changes over time, but it is highly sensitive to spatial changes, so other features that are not related to spatial changes need to be extracted. If it is a color image, more auxiliary features are extracted in channels that can reflect illumination information. The channels include the V channel of HSV, the Y channel of YUV, the L channel of LAB and the weighted average of the three passes of RGB; if it is a grayscale or infrared image, use its single pass value. In this example, the Fourier transform features are extracted from the selected channels of the face area of the two frames of light and dark images, and other suitable features mentioned above may also be used.

As an optional implementation manner, a feature vector is obtained by concatenating the discrete Fourier feature and the auxiliary feature. Using the model of machine learning to build a random forest classifier for living and non-living. The model of this system uses a large number of data (500,000 groups) of human faces and non-human faces collected in different environments (such as human faces on the screen of mobile phones or tablet computers, and photos printed with different materials) for training. The input of the training algorithm is the above feature vector and expected output. Other suitable machine learning algorithms include, but are not limited to, artificial neural networks, decision trees, support vector machines, convolutional neural networks, and the like. Use the score of the classifier to make a judgment. If it is greater than or equal to the threshold, it will be judged as a living body, and if it is less than the threshold, it will be judged as a non-living body.

It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is a relationship between these entities or operations. any such actual relationship or order exists between them. Furthermore, the term "comprises", "comprises" or any other variation thereof is intended to cover a non-exclusive inclusion such that a process, method, article or apparatus comprising a set of elements includes not only those elements, but also includes elements not expressly listed. other elements of or also include elements inherent in such a process, method, article, or device. Without further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article or apparatus comprising said element.

Those of ordinary skill in the art can understand that all or part of the steps in the implementation of the above method can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, referred to herein as Storage media, such as: ROM/RAM, disk, CD, etc.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present invention are included in the protection scope of the present invention.

Claims (24)

1. A human face living body detection system based on light pulse, it is characterized in that, described system comprises: A light-emitting device, the light-emitting device emits light signals of different signal intensities to the object to be measured at a preset frequency; an imaging device, the imaging device is used to acquire multiple images of the object to be measured under light signals of different signal intensities; and A detection device, the detection device detects whether the object to be detected is a living body based on the plurality of images. 2. The system according to claim 1, wherein preferably, the detection device comprises: an extracting device, the extracting device extracts target areas of the object to be measured on multiple images, and acquires feature point information of the target areas; and A judging device, which judges whether the object to be measured is a living body based on the target area and the feature point information. 3. The system according to claim 2, wherein the extraction device is also used for: calculating the variance between the target area and the feature points on the current image acquired by the camera; When the variance is lower than the first threshold, the current image is discarded, and the image of the object to be measured is acquired from the camera device again. 4. The system according to claim 2, wherein the extraction device is also used for: calculating the average brightness value of the target area on each image captured by the camera; Discarding is performed on images whose average brightness value is greater than the second threshold or lower than the third threshold. 5. The system according to claim 4, wherein the extraction device is also used for: Within the range of preset brightness values, the image with the lowest average brightness value of key points is selected as the first image, and the image with the highest average brightness value of key points is selected as the second image. 6. The system according to claim 5, wherein the detection device further comprises: A preprocessing device, configured to perform image preprocessing on the first image and the second image. 7. The system according to claim 6, wherein the detecting whether the object to be measured is a living body based on the plurality of images comprises: The target areas of the first image and the second image are respectively subjected to normalization processing, and the feature points are mapped to a unified standard position. 8. The system according to claim 7, wherein said normalizing said target area and mapping said feature points to a unified standard position comprises: The transformation matrices between the key points of the target area and the predetermined average target area on the first image and the second image are respectively calculated, and the key points of the target area are mapped to standard positions. 9. The system according to claim 7 or 8, wherein the detecting whether the object to be measured is a living body based on the plurality of images further comprises: Difference processing is performed on the normalized first image and the second image to obtain a difference map of the target area. 10. The system according to claim 9, wherein the detecting whether the object to be measured is a living body based on the plurality of images further comprises: Extract discrete Fourier transform features in a difference map: $\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; \&pi;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; l</span>}\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; i</span>}}$ Where e ^ix =cosx+sinx, I(m,n) is the pixel value of the difference map of the target area, and N is the size of the difference map. 11. The system according to claim 10, wherein the detecting whether the object to be tested is a living body based on the plurality of images further comprises: obtain the graphics type of the first image and the second image; and Auxiliary features of the first image and the second image are extracted based on the image type. 12. The system according to claim 11, wherein the detecting whether the object to be tested is a living body based on the plurality of images further comprises: constructing a feature vector using the discrete Fourier transform feature and the auxiliary feature; The feature vector is used as an input, and is analyzed by a classifier based on reflective characteristics established by machine learning to determine whether the target to be tested is a living body. 13. A human face living body detection method based on light pulses, characterized in that the method comprises: Transmit optical signals with different signal strengths to the object to be measured at a preset frequency; acquiring a plurality of images of the object to be measured under light signals of different signal intensities; and Based on the plurality of images, it is detected whether the object to be detected is a living body. 14. The method according to claim 13, wherein the detecting whether the object to be tested is a living body based on the plurality of images comprises: Extracting target areas of the object to be measured on multiple images, and acquiring feature point information of the target areas; and Based on the target area and the feature point information, it is judged whether the object to be measured is a living body. 15. The method according to claim 14, wherein said acquiring a plurality of images of the object to be measured produced under light signals of different signal intensities comprises: calculating the variance between the target area and the feature points on the current image acquired by the camera; When the variance is lower than the first threshold, the current image is discarded, and the image of the object to be measured is acquired from the camera device again. 16. The method according to claim 14, wherein said obtaining a plurality of images of the object to be measured produced under light signals of different signal intensities further comprises: calculating the average brightness value of the target area on each image captured by the camera; Discarding is performed on images whose average brightness value is greater than the second threshold or lower than the third threshold. 17. The method according to claim 16, wherein said acquiring a plurality of images of the object to be measured produced under light signals of different signal intensities further comprises: Within the range of preset brightness values, the image with the lowest average brightness value of key points is selected as the first image, and the image with the highest average brightness value of key points is selected as the second image. 18. The method of claim 17, further comprising: Image preprocessing is performed on the first image and the second image. 19. The method according to claim 18, wherein the detecting whether the object to be tested is a living body based on the plurality of images comprises: The target areas of the first image and the second image are respectively subjected to normalization processing, and the feature points are mapped to a unified standard position. 20. The method according to claim 19, wherein said normalizing said target area and mapping said feature points to a unified standard position comprises: The transformation matrices between the key points of the target area and the predetermined average target area on the first image and the second image are respectively calculated, and the key points of the target area are mapped to standard positions. 21. The method according to claim 19 or 20, wherein the detecting whether the object to be tested is a living body based on the plurality of images further comprises: Difference processing is performed on the normalized first image and the second image to obtain a difference map of the target area. 22. The method according to claim 21, wherein the detecting whether the object to be tested is a living body based on the plurality of images further comprises: Extract discrete Fourier transform features in a difference map: $\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; \&pi;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; l</span>}\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; i</span>}}$ Where e ^ix =cosx+sinx, I(m,n) is the pixel value of the difference map of the target area, and N is the size of the difference map. 23. The method according to claim 22, wherein the detecting whether the object to be tested is a living body based on the plurality of images further comprises: obtaining the graphic type of the first image and the second image; and Auxiliary features of the first image and the second image are extracted based on the image type. 24. The method according to claim 23, wherein the detecting whether the object to be tested is a living body based on the plurality of images further comprises: constructing a feature vector using the discrete Fourier transform feature and the auxiliary feature; The feature vector is used as an input, and is analyzed by a classifier based on reflective characteristics established by machine learning to determine whether the target to be tested is a living body.